Estimating and imputing missing tax loss carryforward data to reduce measurement error

The ability to reduce current and future taxable income with prior years' taxable losses is highly relevant for explaining firms' effective and marginal tax rates. Compustat data on the tax loss carryforward (TLCF) are, however, often missing. We propose a method to impute estimated values for the missing TLCF data instead of the common practice of imputing zero values. In order to assess the accuracy of this method in estimating TLCFs, we compare our imputed values (i) with a random selection of 10-K data, and (ii) with Compustat data for firm-years where Compustat data is available. The results show that our estimated values align very closely with the reported data. Re-analyzing the results of Frank et al. (2009) shows that our imputed values lead to a large decrease in measurement error. This reduces the risk that firms with missing data and low effective tax rates may be incorrectly classified as tax aggressive. A re-analysis of Foley et al. (2007) shows that using our estimated TLCF values may lead to conclusions that are qualitatively different from conclusions based on imputing zero values. Using our imputed values thus increases the probability of correct inferences in studies that use TLCF information from Compustat